Heart Rate During the Mock Audition

performance-anxiety
heart-rate
descriptive
MFA-2026
Author

Brianna Meikle

Published

February 28, 2026

This analysis presents the heart rate data from five mock audition sessions. Each singer performed two songs – one prepared, one assigned 24 hours in advance – for a three-person faculty panel. All subjects wore Polar H10 heart rate monitors. HR was recorded at 1 Hz (one value per second) and aligned to audio using a clap-based sync point.

Participants:

Assigned pieces: Female singers (S2, S3, S4): “If You Knew My Story” | Male singers (S1, S5): “Shiksa Goddess”

Data constraint: HR was exported from Polar Flow at 1 Hz – averaged BPM per second, not beat-to-beat RR intervals. Standard HRV metrics cannot be computed. Analysis focuses on HR magnitude, reactivity, and interpersonal synchrony.

Methodological notes: - All faculty panelists were seated throughout all auditions. No standing, walking, or postural changes occurred. All panelist HR variations reflect genuine autonomic responses, not movement artifacts. - Panelist 1 wore a facial mask for the duration of the experiment to avoid sharing or receiving illness. This may have contributed slightly to Panelist 1’s elevated baseline HR (~100 bpm), though the consistency across all five sessions suggests a stable individual difference rather than a masking artifact.

Part 1: Individual Session Time Series

Each figure shows one audition session with the singer’s HR (red) overlaid with Panelist 1 (purple), Panelist 2 (green), and Panelist 3 (blue). Shaded bands indicate audition phases.

Panelist baseline context: The three panelists – all seated throughout – show a stable hierarchy across all sessions: Panelist 1 runs highest (~98-108 bpm), Panelist 3 is moderate (~75-95 bpm), and Panelist 2 is lowest (~55-65 bpm). Panelist 2’s HR is essentially resting heart rate for a seated adult. These individual differences in resting autonomic tone are consistent across all five sessions, suggesting stable trait-level variation rather than differential responses to specific singers.

Session 1 – Singer 1 (M)

Singer 1 begins around 127 bpm and ramps steadily upward through song 1 (prepared piece), reaching peaks near 160 bpm. There is a brief dip during the inter-song gap, followed by a sharp rise into song 2 (assigned piece), where HR reaches its highest sustained values (~155-158 bpm). The three panelists hold flat: Panelist 1 around 100-108 bpm, Panelist 3 at 85-95 bpm, and Panelist 2 near 58 bpm. The singer’s activation gap above the panelist mean is ~60-70 bpm during performance.

Session 2 – Singer 2 (F)

Singer 2 shows the steepest ramp-up of any singer – HR climbs from ~120 bpm at the start to peak near 169 bpm during song 1 (prepared). HR drops modestly during the inter-song gap (~145 bpm) but rebounds during song 2 (assigned), reaching ~155 bpm. Panelist 1 drifts from 108 down to ~100 bpm; Panelist 3 hovers in the 75-90 bpm range; Panelist 2 holds steady near 57 bpm. The activation gap widens as the session progresses.

Session 3 – Singer 3 (F)

Singer 3 follows a similar arc to Singer 2 – gradual climb through the pre-singing phase, sustained elevation during song 1 (135-165 bpm), partial recovery in the gap, and a plateau during song 2 (~150-158 bpm). Panelist 1 is steady around 101-107 bpm; Panelist 3 is at 80-95 bpm; Panelist 2 stays near 58-60 bpm. The inter-song gap shows a brief but noticeable HR spike in Panelist 3’s trace – likely anticipatory activation as the panelist prepares for the next performance (see Panelist Anticipatory Activation below).

Session 4 – Singer 4 (F)

Singer 4 is a clear outlier. HR starts around 80-90 bpm in the pre-singing phase – within the range typically seen in a panelist, not a singer. During song 1, HR rises modestly to ~105-120 bpm, well below the 140-170 bpm range seen in other singers. During song 2, HR actually drops below baseline, reaching the mid-70s. This is the most visually striking session in the dataset: the singer’s trace overlaps with Panelist 1 and Panelist 3, while Panelist 2 runs below all of them. Panelist 1 spikes to ~113 bpm during the inter-song gap – the highest F1 variability in any session (sd=5.8). Panelist 3 is similarly variable (sd=8.1 during the gap).

Important context: Panelist 3 had a pre-existing private studio teaching relationship with Singer 4. Panelist 1 had previously taught all five participants in a classroom setting. Singer 4’s reduced HR may reflect the familiarity of the studio teacher-student relationship, individual differences in autonomic reactivity, anxiolytic medication use (not screened for), or some combination.

Session 5 – Singer 5 (M)

Singer 5 shows the highest sustained HR of any participant. Starting from a pre-singing baseline of ~135-140 bpm (already elevated), HR climbs rapidly through song 1 to peaks near 184 bpm. The inter-song gap provides minimal recovery (~165 bpm), and song 2 drives HR to its highest sustained values (~170-180 bpm). The three panelists form a clear floor: Panelist 1 at 96-106 bpm, Panelist 3 at 73-85 bpm, Panelist 2 at 54-59 bpm. The widest activation gap in the study occurs here – ~91 bpm between singer and the 3-panelist mean during song 2.

Part 2: Cross-Session Comparisons

Singer HR by Audition Phase

The grouped bar chart below shows each singer’s mean HR (with standard deviation bars) during the four performance phases. The pattern is consistent across four of five singers: HR is lowest during pre-singing, rises sharply for song 1, dips slightly during the inter-song gap, and remains elevated or continues rising into song 2. Singer 4 is the exception – her HR stays in the 88-106 bpm range across all phases, never approaching the 140+ bpm levels seen in the others.

Key observation: Pre-singing baseline HR is remarkably similar across Singers 1, 2, 3, and 5 (~126-143 bpm), suggesting comparable levels of anticipatory anxiety before singing begins. Singer 4’s pre-singing HR (~91 bpm) is 35-50 bpm lower than the group.

Singer-Panelist Activation Gap

The chart below quantifies the HR difference between each singer and the mean of all three panelists for each phase. The gap reflects how far the singer’s autonomic activation exceeds the panelists’ – a rough index of the anxiety asymmetry in the audition relationship. With Panelist 2 included (mean HR ~58 bpm, seated), the 3-panelist average is lower than the earlier 2-panelist estimate, so activation gaps are wider.

Singers 1, 2, 3, and 5 show gaps of 38-91 bpm, widening as the session progresses. The gap is largest during the inter-song gap and song 2 for most singers – suggesting that the stress of performing the assigned piece (less familiar, less rehearsed) compounds the already elevated state from song 1.

Singer 4’s gap is small but positive (+8 bpm during song 2) – her HR stays close to the panelist range. This is the only singer whose physiological state approximates the panelists’.

HR Reactivity Profiles

The dot-and-line plot below connects each singer’s mean HR across the four phases, making the shape of each stress response visible.

Four distinct profiles emerge:

  • Singer 5 – highest absolute HR, steep climb, no recovery between songs. The most pronounced stress response in the study.
  • Singer 2 – steep initial climb during song 1, partial recovery in the gap, rebound for song 2. Classic performance anxiety arc.
  • Singers 1 and 3 – moderate, steady climb. Singer 1 shows the most gradual escalation; Singer 3 plateaus earlier.
  • Singer 4 – flat trajectory near 90 bpm. Modest rise during song 1, return to baseline for song 2. This profile resembles a panelist more than a performer.

Normalized HR Trajectories

All five singers’ HR traces plotted on a common timeline, each zeroed to their own pre-singing baseline mean. This removes differences in resting HR and shows the magnitude of change from baseline.

With a 10-second rolling average for readability:

  • Singers 2 and 5 show the largest sustained elevations (~25-40 bpm above baseline).
  • Singer 1 shows a moderate, steady rise (~15-30 bpm above baseline).
  • Singer 3 shows a moderate rise with more variability.
  • Singer 4 shows the smallest reactivity – peaking around +15 bpm during song 1, then dropping below baseline during song 2.

Part 3: Same-Song Comparisons

Because the assigned piece was the same within each gender group, we can compare singers’ HR during the identical musical material. Any differences in HR trajectory during the same song reflect individual anxiety responses rather than differences in musical demands.

Female Singers – "If You Knew My Story"

During the assigned piece (right panel) – the same 32-bar cut – Singers 2 and 3 track closely in the 140-155 bpm range with similar contour, while Singer 4 starts at ~80 bpm and rises slowly to ~99 bpm. The separation between Singer 4 and the other two is even more striking here than during the prepared pieces, because the musical demands are identical.

Each female singer during "If You Knew My Story" with all three panelists overlaid:

  • Singer 2: Sustained at 140-155 bpm. All three panelists hold flat below, with Panelist 2 running lowest (~57 bpm). The singer sits 40-80+ bpm above the panelist range.
  • Singer 3: Similar profile to Singer 2. Panelist traces are stable and well below.
  • Singer 4: Starts at ~80 bpm – below Panelist 1 (~98 bpm) and near Panelist 3 (~92 bpm). Only Panelist 2 (~56 bpm) is consistently below her. This is the only session where the singer’s HR is intertwined with the panelists’ during active singing.

Male Singers – "Shiksa Goddess"

During the assigned piece (right panel), Singer 5 starts at ~166 bpm and climbs to ~180 bpm. Singer 1 starts at ~149 bpm and rises gently to ~158 bpm. The ~15-20 bpm gap between them is consistent, suggesting a stable difference in stress reactivity rather than a response to a specific musical moment.

Each male singer during "Shiksa Goddess" with all three panelists:

  • Singer 1: Flat trace at ~150-158 bpm. All three panelists well below. The gap between singer and 3-panelist mean (~70 bpm) is stable throughout.
  • Singer 5: Climbing trace from ~166 to ~180 bpm. The gap widens over the course of the song – panelists hold flat or decrease while Singer 5 continues climbing. Panelist 3 shows the clearest antiphase pattern (HR decreasing as singer’s increases).

Part 4: Interpretive Notes

The Singer 4 - Panelist 3 Relationship

Singer 4 is a student of Panelist 3 – they have a pre-existing private studio teacher-student relationship. Panelist 1 also had prior classroom teaching experience with all five participants, though in a group setting rather than one-on-one. These relational contexts are important interpretive lenses, but cannot be isolated as causal explanations for Singer 4’s low HR. Several factors may contribute:

  1. Singer 4’s low HR is consistent with reduced evaluative threat. Performing for one’s own studio teacher – someone who knows your voice, has heard you at your worst, and is invested in your success – may reduce the sense that resources are insufficient to meet demands (the "threat" state in the biopsychosocial model; Guyon et al., 2020). However, the other singers also performed for a familiar classroom teacher (Panelist 1) without a comparable reduction, suggesting that the studio relationship’s intimacy may matter, or that individual differences in autonomic reactivity or coping style contribute independently.

  2. The HR convergence between Singer 4 and Panelist 3 is consistent with what Coutinho et al. (2021) describe in established relationships. While Singer 4 and Panelist 3 are not romantic partners, the private studio relationship represents the closest interpersonal bond of any singer-panelist dyad in this study, and it produces the most convergent physiological profile.

  3. All three panelists show elevated variability during Singer 4’s session. Panelist 3’s HR is the most variable of any session (sd=8.1 during the inter-song gap). Panelist 1 also shows its highest variability here (sd=5.8, spiking to ~113 bpm). Even Panelist 2, typically the most stable, shows slightly more variation.

For the thesis: This dyad should be discussed as both a limitation (the pre-existing relationships are confounds) and as a case study in co-regulation (the strongest evidence of physiological coupling occurs in the dyad with the deepest relational history). The explanation should be framed as one plausible factor rather than a confirmed mechanism, given that individual differences in autonomic reactivity or anxiolytic medication use (not screened for) could also account for Singer 4’s profile.

Anticipatory Anxiety

Four of five singers show pre-singing HR values of 126-143 bpm – substantially elevated above typical resting heart rate (60-80 bpm for young adults). This is consistent with Vellers et al. (2017), who identified the pre-performance anticipatory period as the most sensitive window for detecting audition stress. The singers are already in a state of sympathetic activation before they begin singing. Singer 4, again, is the exception at ~91 bpm.

Panelist Anticipatory Activation

The singer anticipatory anxiety finding has a counterpart on the panelist side. All three panelists – seated throughout – show HR ramp-ups in the final seconds of the inter-song gap before song 2 begins. Because the panelists did not stand, shift position, or move during the auditions, these HR changes cannot be attributed to postural artifacts. They reflect genuine autonomic activation – the evaluative experience of preparing to hear the next piece.

Panelist 3 shows the most consistent pattern (anticipatory activation in 4 of 5 sessions), with Session 4 (her own student) producing the largest ramp (+17.7 bpm). Panelist 1 shows activation in 3 of 5 sessions. Panelist 2, with the lowest overall HR, shows modest changes that are harder to distinguish from baseline fluctuation.

Prepared vs. Assigned Song

The data does not show a simple pattern of "assigned piece = more anxiety." Some singers (1, 3, 5) show higher HR during song 2 (assigned), but Singer 2 shows lower HR during song 2 than song 1. This may reflect:

  • Fatigue or habituation (HR naturally declining after sustained high effort)
  • The assigned piece being shorter and less demanding than some prepared pieces
  • Individual differences in how novelty vs. familiarity affects stress

The comparison is also confounded by order – song 2 always follows song 1, so any "assigned piece" effect is inseparable from a "second performance" effect. This should be acknowledged as a design limitation.

K-MPAI and MAAQ Pre-Survey Results

The K-MPAI was administered to all BoCo BFA/MM voice students as a pre-screening survey. The 5 study participants were drawn from this broader pool of 26 respondents. Scoring follows Kenny (2009): 40 items on a 0-6 scale with 8 reverse-coded positive items (1, 2, 9, 17, 23, 33, 35, 37), range 0-240, clinical threshold >=105.

Population (26 BoCo singer respondents):

  • Mean K-MPAI: 124.1 (SD 31.6), Median: 124.0, Range: 64-185
  • 73% (19/26) score >=105 (clinically significant MPA)
  • This is a high-anxiety population, consistent with conservatory performance culture

Study participants:

Singer K-MPAI Clinical? MAAQ Flexibility Notes
S1 (M) 126 Yes 31 Near population mean
S2 (F) 111 Yes 35 (highest) Lowest anxiety, highest flexibility
S3 (F) 144 Yes 26 Above population mean
S4 (F) 159 Yes 21 High anxiety despite lowest HR
S5 (M) 174 Yes 17 (lowest) Highest anxiety, lowest flexibility

All 5 singers score above the clinical threshold – not surprising given they self-selected for a study on performance anxiety. The rank ordering is notable: Singer 5 (highest K-MPAI, 174) also shows the highest sustained HR in the study, while Singer 2 (lowest K-MPAI, 111) shows the steepest HR ramp-up but also the most recovery between songs. Singer 4 (K-MPAI 159 – second highest) is the most interesting case: high self-reported anxiety with the lowest physiological response. Several factors may contribute, including the studio relationship with Panelist 3, individual autonomic differences, or anxiolytic medication (not screened for in this study).

Faculty panelists:

Panelist K-MPAI Clinical? MAAQ Flexibility Mean HR (seated)
Panelist 1 109 (2 items missing) Yes 24 ~101 bpm
Panelist 2 102 No 33 ~58 bpm
Panelist 3 169 Yes 32 ~85 bpm

Panelist 3’s K-MPAI (169) is notably high – higher than any study participant except Singer 5. For a panelist whose own performance anxiety is elevated, evaluating singers (especially her own student, Singer 4) may carry additional physiological weight. Panelist 2 falls just below the clinical threshold at 102 and shows the lowest HR of anyone in the study – essentially resting heart rate while seated. The K-MPAI ranking on the faculty side (F3 > F1 > F2) does not map directly to their HR ranking (F1 > F3 > F2), but Panelist 2 being the calmest on both measures is consistent.

Limitation: The study did not screen for anxiolytic medication use (beta blockers, SSRIs, benzodiazepines) in singers or faculty panelists. Any participant using such medication could show attenuated HR responses that do not reflect their underlying anxiety level.

K-MPAI Scoring Note

The Qualtrics-generated SC0 score differs from our computed score by a constant of +40 across all 26 respondents (verified: identical SDs, perfect rank correlation). Qualtrics scores the K-MPAI on a 1-7 internal scale, while Kenny (2009) specifies 0-6. Both apply the same reverse coding on the same 8 items – the only difference is a +1 shift per item across all 40 items (40 x 1 = 40). We use the published 0-6 convention. The Qualtrics threshold equivalent of >=105 is >=145.

Summary

Singer Pre-Singing HR Song 1 HR Song 2 HR Reactivity (Song 1 - Baseline) Activation Gap (Song 2) K-MPAI Notes
S1 (M) 128 144 154 +16 +70 126 Steady climber; flattest assigned-song trace
S2 (F) 128 160 147 +32 +66 111 Steepest ramp-up; HR drops for song 2
S3 (F) 126 150 152 +24 +70 144 Moderate, steady; similar to S2 during assigned song
S4 (F) 91 106 90 +15 +8 159 Outlier – pre-existing relationship with F3; HR near panelist range
S5 (M) 143 167 170 +24 +91 174 Highest absolute HR; no recovery between songs

All values are mean HR in bpm. Activation gap = singer mean HR minus mean of all three panelists during song 2.

Part 5: Anxiety Measures x Heart Rate Correlations

K-MPAI x Heart Rate (RQ2)

Does higher trait performance anxiety predict higher physiological arousal during the audition? With n = 5 singers, Spearman rank correlations are used. The critical value for significance at p < 0.05 (two-tailed) with n = 5 is |r_s| = 1.000, so no correlation can reach conventional significance. These are descriptive rank-order patterns, not inferential tests.

Anxiety Measure HR Measure r_s p
K-MPAI Total Mean Song HR +0.10 0.873
K-MPAI Total Peak HR +0.10 0.873
K-MPAI Total HR Reactivity -0.30 0.624
K-MPAI Total Activation Gap +0.10 0.873
MAAQ Flexibility Mean Song HR -0.10 0.873
MAAQ Raw Sum Mean HR (Assigned Song) -0.87 0.054

No meaningful K-MPAI-HR association. Singer 4’s position breaks any monotonic pattern: she has the second-highest K-MPAI (159) but the lowest HR across every measure. If Singer 4 were removed, the remaining four singers show a near-monotonic positive relationship between K-MPAI and Mean Song HR. Several factors may contribute to Singer 4’s low HR: the pre-existing studio teaching relationship with Panelist 3, individual differences in autonomic reactivity, or anxiolytic medication use (not screened for in this study). With n = 5, a single case can shift the entire correlation.

Post-Survey x Heart Rate Correlations

Singer Self-Report

Singers completed an 8-item post-performance survey (1-5 Likert) immediately after their audition. Each item is correlated with their mean song HR.

Critical finding: “Felt anxious” has zero variance. All five singers rated their anxiety at 4 out of 5. With no rank-order variance, the correlation is undefined. This is itself a substantive result: every singer, including Singer 4 (whose HR was near panelist levels), self-reported high anxiety. The audition context produced a universal subjective experience of anxiety regardless of the physiological response, consistent with research showing that cognitive and somatic anxiety components can dissociate.

Survey Item Direction vs Mean Song HR (r_s)
Felt confident positive +0.35
Felt anxious negative NaN (no variance)
Felt authentic positive 0.00
Felt safe to express positive 0.00
Emotionally connected positive +0.78
Emotionally disconnected negative -0.05
Enjoyed performing positive +0.37
Did not feel satisfaction negative +0.63

“Emotionally connected” shows the strongest positive correlation with HR (r_s = +0.78, p = 0.118). Singers with higher HR reported feeling more emotionally connected to their performance, not less. This is consistent with the idea that physiological activation during performance is not purely aversive: it can coexist with engagement and emotional presence.

Faculty Perception x Singer HR

Faculty panelists rated each singer on 6 items after each audition (n = 15 observations: 3 panelists x 5 singers, not independent). These ratings were correlated with both singer HR and the panelist’s own HR during that session.

Faculty Item vs Singer Mean HR (r_s) vs Panelist Own HR (r_s)
Appeared confident +0.50 (p=.057) -0.14
Showed signs of anxiety -0.40 (p=.138) -0.12
Felt authentic -0.03 -0.20
Emotionally connected -0.07 -0.03
Emotionally disconnected -0.09 +0.12
Appeared to enjoy performing +0.34 +0.02

Panelists rated singers with higher HR as appearing more confident, not less (r_s = +0.50, p = 0.057). And they rated singers with higher HR as showing fewer signs of anxiety (r_s = -0.40). This is driven by Singer 4 (lowest HR) receiving some of the highest anxiety ratings, while Singer 5 (highest HR) received the lowest anxiety ratings from two of three panelists. Physiological stress was not visible to the panelists.

Panelist own HR shows no meaningful correlation with any perception item. What panelists felt physiologically did not predict how they rated the singer.

Data Pending

  • Exploratory analyses (Phase 4) – individual case studies, panelist convergence, audition order effects.

Part 6: HR Trajectory Features and Individual Growth Curves

The mean-level correlations in Part 5 collapse each singer’s entire song performance into a single HR value. This section preserves temporal structure by extracting trajectory features from HR time series during each song phase and fitting individual growth curves.

Trajectory Features

Five features were extracted from each singer’s HR during song_1 and song_2, then averaged across both songs:

  • Slope: linear regression slope of HR over normalized time (0 = song onset, 1 = song offset). Using normalized time avoids a confound with song duration.
  • AUC above baseline: area under the HR curve above the singer’s pre-singing baseline HR, normalized by duration (bpm above resting).
  • Time to Peak: seconds from phase start to maximum HR
  • HR Range: max minus min HR within the phase (amplitude of fluctuation)
  • Recovery: HR at phase end minus HR at phase start (positive = continued climbing, negative = recovery)
Feature K-MPAI r_s p MAAQ Flex. r_s p
Slope (bpm/norm. time) +0.70 0.188 -0.70 0.188
AUC above baseline (bpm) -0.30 0.624 +0.30 0.624
Time to Peak +0.30 0.624 -0.30 0.624
HR Range +0.90 0.037 -0.90 0.037
Recovery +0.70 0.188 -0.70 0.188

Key findings:

  1. HR Range is the strongest trajectory-anxiety association (r_s = +0.90, p = 0.037, uncorrected). Higher trait anxiety predicts wider HR fluctuation during singing. Slope and Recovery also show positive associations with K-MPAI (r_s = +0.70 each) but do not reach conventional significance. These trajectory features reveal information that mean HR obscures. Unlike the near-zero K-MPAI x Mean Song HR correlation (r_s = +0.10), the trajectory features capture the dynamic quality of the stress response.

Note: All p-values are uncorrected for multiple comparisons. Across the full analysis, over 50 correlations are computed. Even the one p = 0.037 result (HR Range) would not survive Bonferroni correction. All correlations should be treated as exploratory pattern descriptions.

  1. Singer 4 no longer breaks the pattern for HR Range. In the mean HR analyses, Singer 4’s low absolute HR (despite high K-MPAI) produced flat correlations. But Singer 4’s range is actually consistent with moderate-to-high K-MPAI: she shows meaningful HR fluctuation (23-40 bpm range) even though her absolute HR stays low. The trajectory features reveal that Singer 4’s autonomic system is responding to the performance, just from a lower set point.

  2. Recovery is positive for all singers. No singer shows net HR recovery within either song phase. All singers’ HR is still climbing or elevated at the end of each song.

Individual Growth Curves

For each singer and song phase, HR was modeled over normalized time (0 = song onset, 1 = song offset) using both linear and quadratic regression. The quadratic fit captures inverted-U or U-shaped trajectories that a linear model would miss.

Singer Song Shape Linear R2 Quad R2 Notes
S1 Prepared Ramp-up 0.852 0.861 Steady linear climb
S1 Assigned Ramp-up 0.733 0.736 Steady linear climb
S2 Prepared Inverted-U 0.044 0.913 Classic arc: rise then fall
S2 Assigned Ramp-up 0.353 0.384 Shifts to climb pattern for new piece
S3 Prepared Inverted-U 0.739 0.794 Rise with late plateau
S3 Assigned Inverted-U 0.418 0.599 Moderate arc
S4 Prepared Inverted-U 0.013 0.428 Rise and fall within panelist range
S4 Assigned Ramp-up 0.889 0.889 Pure linear climb from low baseline
S5 Prepared Inverted-U 0.797 0.894 Steep climb with late plateau near 180 bpm
S5 Assigned U-shape 0.799 0.902 Brief early dip then steep climb

Interpretive notes:

  1. S2 prepared piece is the most dramatic inverted-U (R2 improvement = +0.87). A linear model explains almost nothing (R2 = 0.04), while the quadratic captures 91% of variance.

  2. The prepared-to-assigned shift changes trajectory shape. Three singers (S2, S4, S5) show different shapes between songs. The assigned piece may not allow enough time or familiarity for the downslope of the inverted-U to emerge.

  3. S1 is the most consistently linear. Both songs show high linear R2 (0.73-0.85) with negligible quadratic improvement.

Part 7: Physiological Synchrony (Windowed Cross-Correlation)

This section quantifies the degree to which singer and panelist HR move together during performance, using windowed cross-correlation (WCC). This addresses RQ1: does physiological coupling (synchrony) occur between singer and panelist during the audition?

Method

Following Moulder et al. (2018), WCC was computed with: - Window: 10 seconds (captures meaningful HR fluctuation at 1 Hz) - Step: 5 seconds (50% overlap) - Max lag: +/-8 seconds (physiological response latency) - Primary metric: mean detrended Pearson r at lag 0

Within each window, both signals are linearly detrended before correlation. Windows where either signal has SD < 0.5 bpm are excluded (the “near-constant guard,” primarily affecting Panelist 2).

Surrogate testing: Each singer’s HR was paired with each panelist’s HR from other sessions (session-swap). This preserves temporal structure while breaking real coupling.

Panelist 2 Exclusion Rate

Panelist 2’s HR was near-constant in 95% of windows (188 of 198 excluded). Panelist 2 cannot contribute meaningful synchrony data, confirming the near-resting physiological state observed in Parts 1-4.

WCC Results

Singer Panelist 1 (mean r) Panelist 3 (mean r) Panelist 2 (mean r) Notes
S1 +0.23 +0.23 +0.04 Moderate positive synchrony
S2 +0.19 -0.08 -0.36 Positive with Panelist 1 only
S3 +0.07 -0.01 +0.03 Near zero across all panelists
S4 +0.36 +0.03 +0.45 Strongest Panelist 1 coupling
S5 +0.03 -0.15 NaN Panelist 3 shows antiphase pattern

Key observations:

  1. Singer 4 shows the strongest synchrony, but with Panelist 1, not Panelist 3. The S4-Panelist 1 pair has the highest mean r across both songs (+0.36), with particularly strong coupling during song_2 (r = +0.55). The studio teacher-student pair (S4-Panelist 3) shows near-zero coupling (+0.03).

  2. Singer 5 and Panelist 3 show the only consistent antiphase pattern (mean r = -0.15). Panelist 3’s HR tends to move opposite to Singer 5’s during performance.

  3. Most real pairs do not clearly exceed surrogates. With n = 5 and short song durations, statistical power is low. This question remains open for larger studies.

Synchrony and K-MPAI

Comparison r_s p
K-MPAI vs Mean Sync (all panelists) -0.10 0.873
K-MPAI vs Sync with Panelist 1 -0.30 0.624
K-MPAI vs Sync with Panelist 3 -0.30 0.624
MAAQ Flex vs Mean Sync +0.10 0.873

No meaningful relationship between trait anxiety and physiological synchrony. The direction is weakly negative (higher K-MPAI = slightly less synchrony), but the magnitudes are negligible.

Synchrony by Condition: Prepared vs. Assigned Song

The paired dot plot below shows each singer-panelist pair’s synchrony during Song 1 (prepared) vs. Song 2 (assigned). No consistent pattern emerges: some pairs show higher synchrony during the prepared piece, others during the assigned piece. The short duration of Song 2 (~55-59 seconds) limits the number of WCC windows and makes these estimates noisier.

Part 8: Trend-Level Cardiac Coherence (Smoothed Whole-Series Correlation)

The WCC analysis (Part 7) detrends each 10-second window before computing correlations, targeting moment-to-moment fluctuation coupling. However, visual inspection of the HR time series suggests a broader pattern: singer and panelist HR traces often share the same general contour over the course of a song, rising and falling together at a timescale of 15-60 seconds. The WCC detrending removes exactly this kind of slow co-movement.

To test whether this trend-level coherence is real, both HR signals are smoothed with a 15-second centered moving average, then a single Pearson correlation is computed on the full smoothed series for each singer-panelist-song pair.

Pair Song 1 r Song 2 r Mean r
S1-F1 +0.391 -0.838 -0.224
S1-F3 +0.537 -0.940 -0.202
S2-F1 -0.860 +0.857 -0.002
S2-F3 +0.644 +0.897 +0.771
S3-F1 +0.462 +0.857 +0.660
S3-F3 +0.541 +0.337 +0.439
S4-F1 -0.168 +0.068 -0.050
S4-F3 +0.297 +0.474 +0.386
S5-F1 +0.506 +0.270 +0.388
S5-F3 +0.405 +0.496 +0.451

F2 omitted from table (near-constant HR). Mean smoothed r across all 30 pairs: +0.220. Range: -0.940 to +0.897.

Seven of 30 pairs exceed 95% of session-swap surrogates. K-MPAI vs Mean Smoothed r: rs = +0.000, p = 1.000. Trend-level coherence is not associated with anxiety level, suggesting shared environmental response rather than anxiety-driven coupling.

Smoothed Correlation by Condition: Prepared vs. Assigned Song

Part 9: Wavelet Coherence

Wavelet coherence decomposes both HR signals into time-frequency space using a continuous Morlet wavelet transform, measuring coherence (0 to 1) at each combination of time and period (timescale). This reveals where in the song and at what timescale the coupling is strongest.

Periods analyzed: 6 to 58 seconds (limited by song duration). The coherence heatmaps show coherence concentrated at longer timescales (30-58s periods), confirming that singer-panelist coupling is a slow, broad phenomenon.

Pair Song 1 Coh. Song 2 Coh. Mean Coh.
S1-F1 0.557 0.297 0.427
S1-F3 0.469 0.312 0.391
S2-F1 0.443 0.394 0.419
S2-F3 0.605 0.399 0.502
S3-F1 0.706 0.651 0.679
S3-F3 0.554 0.642 0.598
S4-F1 0.435 0.466 0.451
S4-F3 0.487 0.454 0.471
S5-F1 0.514 0.388 0.451
S5-F3 0.511 0.412 0.462

Mean coherence across all 30 pairs: 0.489. Six of 30 pairs exceed 95% of surrogates. S3-F1 song_1 shows the highest coherence (0.706, z = +3.95). K-MPAI vs Wavelet Coherence: rs = -0.100, p = 0.873.

How the Three Synchrony Analyses Relate

Analysis Timescale What It Detects Key Finding
WCC (Part 7) 5-10s fluctuations Moment-to-moment coupling after detrending Mostly null; few pairs exceed surrogates
Smoothed r (Part 8) 15s+ trends Broad HR trajectory co-movement Strong coherence in many pairs (mean r = +0.22)
Wavelet (Part 9) 6-58s decomposed Multi-scale coherence, localized in time Moderate coherence concentrated at 30s+ periods

The three analyses converge: singer and panelist HR share broad arcs over the course of a song, but this shared trajectory is not driven by anxiety level and does not extend to moment-to-moment fluctuation coupling.